concept
active
concept:conditional-mean-score-improvement-metricConditional Mean Score Improvement (metric)
Score delta between last and first attempt for multi-attempt responses, measuring correction effectiveness
Neighborhood — ranked by edge-count
Concepts (1)
concept
- Endogenous Steering ResistanceimplementsThe central phenomenon introduced by this paper: inference-time recovery from irrelevant activation steering in LLMs
Conceptual bridges
2-hop · via this concept's ideasWhere ideas in this concept connect to the rest of the corpus — the same concept, an analogy, or a restatement elsewhere.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Metric averaged over all tasks to measure MTL method improvement over STL.
- The increase in reward during training, whose dynamics align with those of causal emergence in successful agents.
- Primary performance metric: total food visits across agent lifetime
- Clarifies nature of S.
- Encoding of prediction confidence; proposed role for dopamine beyond reward signalling.
- The direction of information increase is relative to the observer or user of the computationclaim0.720Example: 3×5→15 is a natural computation, but 15→3×5 (prime factorization) is also useful, showing that the 'gain' depends on the choice of normal form.
- Secondary metric: percentage of responses containing multiple attempts, separating surface from actual self-correction
- Used for updating hidden state expectations; provides dynamical process theory testable against neuronal data